 Correspondence to : iwayansunarya@gmail.com Received: February, 2019 Accepted: March, 2019 Published: March, 2019 JURNAL AKUNTANSI, MANAJEMEN DAN EKONOMI Vol. 21, No. 1, 2019, pp. 32-45 Published online in http://jos.unsoed.ac.id/index.php/jame ISSN: 1410-9336 / E-ISSN: 2620-8482 INTRODUCTION Banks as financial institutions are business entities that collect funds from the community in the form of deposits and then redistribute the funds to the public in the form of loans for a certain period of time. Activities to raise funds greatly determine the amount of funds that can be developed by banks planting funds that generate income for the bank (Nazrian and Hidayat, 2012). According to Faradila and Cahyati (2013), the emergence of banks and financial institutions for conventional banks has been applied in Indonesia. So that Islamic experts form Islamic banks which according to law No.10 of 1998 concerning Islamic banks are banks that carry out their business activities based on Syariah principles which in their activities provide services in payment traffic. Syariah principles according to Article 1 paragraph 13 of Law No. 10 of 1998 concerning banking is an agreement based on Islamic law between banks and other parties for depositing funds or financing business activities, or other activities declared in accordance with Syariah, including financing based on the principle of profit sharing (mudharabah), financing based on equity principles (musyarakah ), the principle of buying and selling goods with profit (murabahah), or financing capital goods based on the principle of pure rent without choice (ijarah), or by the option of transferring ownership of goods leased from the bank by another party (ijarah wa iqtina). This Islamic bank is one of the most sought after banks by Indonesian people who are Muslim because Islamic banks adhere to Islamic principles in Islam. Islamic banks do not apply the interest system but implement a profit-sharing system, namely a fund management system in the Islamic economy. The calculation of profit sharing is based on the consensus of the bank and the customers who invest their funds in Islamic banks. The amount of the customer's right to the bank in calculating the profit sharing is determined by a ratio number or the amount of the part called Nisbah (Sari, 2016). Just like conventional banks, Islamic banks must maintain the performance of their financial statements. Analysis of the performance of financial statements can be seen from the level of financial statements between the The Effect Of Capital Adequacy Ratio, Efficiency And Liquidity On Rentability In Syariah Banks Owned By The Indonesia Government From 2009 - 2017 I WAYAN SUNARYA1 1STIMIK STIKOM, Indonesia Abstract The ratio of financial statements to Islamic banks is one of the determining factors in financial health within the bank itself. For this reason, it is necessary to analyze the influence of capital adequacy, efficiency and liquidity on profitability in Indonesian government-owned Islamic banks from 2009-2017. This study aims to model the effect of capital adequacy (CAR), Efficiency (OEOI) and Liquidity (FDR) on Rentability (ROA), then analyze the model, and provide forecasting and structural analysis of the model. Therefore, the method used in this study is the analysis of Vector Error Correction Model which is applied to time series data from the level of CAR, OEOI, FDR to ROA. Based on the specification, estimation and examination of the model, the VECM(2) model was obtained as the best model. The results of the model analysis say that there is a long-term and short-term causality relationship between the levels of CAR, OEOI, FDR against ROA. Then, based on forecasting and structural analysis, it can be concluded that the results obtained are accurate. Keywords Capital Adequacy Ratio, Return On Asset, Operational Costs for Operational Income, Finance to Deposit Ratio JURNAL AKUNTANSI, MANAJEMEN DAN EKONOMI, VOL 21, N0 1, 2019, 32-45 level of capital adequacy (Capital Adequacy Ratio), the level of profitability or return on assets (ROA), efficiency or OEOI (operational costs against historical income) and the level of liquidity or Finance to Deposit Ratio ( FDR). According to Nur Gilang Giannini (2013) said that in addition to the available funds, the supply of bank credit was also influenced by bank perceptions of the debtor's business prospects and the condition of the banking system itself, such as capital (CAR), the amount of bad loans (NPL), and Loan to Deposit Ratio (LDR), besides that the profitability factor or the level of profit reflected in Return on Assets (ROA) also affects bank credit. For data on ROA of Indonesian government Islamic banks from 2009 to 2017, Bank BTN Syariah has the highest rate in 2017 of 1.71% compared to other Islamic banks. This means that the BTN Syariah Bank in generating net profit after tax comes from assets owned by 1.71%. The next level of data is the level of capital adequacy. In maintaining this level of capital adequacy, Bank Indonesia issued regulation No.6/10/PBI/2004, about the Commercial Bank Soundness Rating System. Provisions which one of them is regulating capital Minimum Bank (Capital Adequacy Ratio) of 8% (eight percent). Based on government regulations regarding bank soundness general, each bank strives to maintain the value of its Capital Adequacy Ratio in order to maintain the soundness of the Bank. In the implementation of the Indonesian government Syariah Bank remains experiencing fluctuations in the development of capital adequacy. From the value of the Indonesian government's Capital Adequacy Ratio Syariah period 2009-2017, BRI Syariah Bank where in 2009, BRI Syariah Bank had a CAR value of 17.04% then in 2017 BRI Syariah Bank had a CAR value of 20.63%, this means there is an increase in BRI Syariah Bank CAR value of 3.59%. The third level, which is the level of operational costs for historical income (OEOI), is the level of comparison of operational costs against historical income. This level is used to measure the level of efficiency and ability of banks to carry out their operations, especially credit. From the value of the level of operational costs against operational income (OEOI) of the Indonesian government Syariah Banks from 2009-2017 observation period, the highest level of OEOI is BRI Syariah Bank where the OEOI level in 2009 was 97.50% to 95.24%. The next level, which is Finance to Deposit Ratio (FDR), states how far the bank is able to repay funds withdrawals made by depositors by relying on financing provided as a source of liquidity. Based on the description above, the condition of the growth of Islamic banks which is faster than conventional banks requires research on the health of banks, one of which is using indicators of capital adequacy. Therefore, the objectives to be achieved in this study is analyzing the effect of capital adequacy, efficiency and liquidity on profitability of the Syariah Bank owned by the Indonesian government from 2009-2017. This study also examine the relationship of capital adequacy, efficiency and liquidity in the run short and long term to the profitability of the Indonesian government Islamic banks from 2009-2017. Bank Based on Law No. 7 of 1992 as amended by Law No. 10 of 1998 concerning banking states: "Banks are business entities that collect funds from the public in the form of deposits and distribute them to the public in order to improve the lives of many people". In terms of benefits or services for the use of funds, both bank deposits and loans can be divided into two, namely (Wahyu, 2016) conventional bank and syariah bank. Conventional banks, which are banks whose activities, both fund raising and fund distribution, provide and impose rewards in the form of interest or a number of rewards in the percentage of funds for a certain period. Syariah Bank, which is a bank that is in its activities, both collection funds and distribution of funds providing and imposing benefits on the basis of Syariah principles, namely buying and selling and profit sharing. Islamic banks are banks that operate without relying on interest. Bank Syariah can also be interpreted as an optimal financial / banking institution and its products are developed based on the Al-Quran and hadith. There are two meanings, namely Islamic banks and banks operating under the principle of Islamic law. Islamic banks are banks that operate with Islamic law and the procedure for its operation refers to the provisions of the Koran and hadith (Wibisono, 2017). JURNAL AKUNTANSI, MANAJEMEN DAN EKONOMI, VOL 21, N0 1, 2019, 32-45 Return On Assets (ROA) According to Bilian and Purwanto (2017), ROA is a comparison between profit after tax and total assets in a period. This level can be considered as a measure of financial health. This level is very important and is related to the performance of the bank because we can see the level of business efficiency of a bank from the profits obtained by using its assets. In the framework of the bank's health assessment, BI will give a maximum value of 100 (healthy) if the bank has an ROA of> 1.5%. So that from the statement, the formula in calculating Return on Assets is as follows: Earning Before Tax ROA = x 100% Average of Total Asset Capital Adequacy Ratio (CAR) According to Bernardin (2016), the Capital Adequacy Ratio is a level that shows how much the total bank assets that contain risk (credit, participation, securities, bills on other banks) are also financed from and bank's own capital in addition to obtaining funds - dana from sources outside the bank, such as public funds, loans, and so on. According to the Bank Indonesia Circular No. 6/23/DPNP dated May 31, 2004 CAR is formulated as follows: Owner's equity CAR = x 100% Weighted Assets at Risk Operational Costs for Operational Income The success of banks is based on a quantitative assessment of bank profitability can be measured using the level of operational costs against historical income. The level of operational costs is used to measure the level of efficiency and ability of banks to carry out their operational activities. The level of Operational Costs to Operating Income is often called the level of efficiency used to measure the ability of bank management to control operational costs against historical income. The smaller this level means the more efficient operational costs incurred by the bank concerned (Fadjar, et. al., 2013). Financing to Deposit Ratio (FDR) According to Wibisono (2017) states Financing to Deposit Ratio (FDR) is a tool to measure the extent of the ability of banks to pay depositors' withdrawals which direct funds have been channeled by the bank to the community by means of loans. FDR will show the Bank's ability to channel third party funds collected by the Bank concerned. The Financing to Deposit Ratio formula (Wahyu, 2016) is as follows: Total Financing FDR = x 100% Total Funds Unit Root Test According to Sinay (2014), the VECM model is based on data time series that are not stationary but are cointegrated. To check stationary data, a unit root test can be used, with the test statistic used is Augmented Dickey-Fuller (ADF). The formula used to analyze the unit root test using the Augmented Dickey-Fuller formula (Moroke, et. al., 2014) as follows: k t 0 t-1 i t-i t i=1 ΔY = α + β Y + β ΔY + ε Where Δ is the first distinguishing operator; t is time; k presents the number of lags used and ε is the error rate; α and β are limiting models. For the Augmented Dickey-Fuller test includes a constant trend and time. According to Hamilton (1990) for the process of receiving and rejecting hypotheses in the unit root test using Augmented Dickey-Fuller using the assumption that the series follows the autoregressive process by accepting and rejecting the null hypothesis (H0) based on regression analysis: p-1 t i t-1 j t-j t j=1 Z = μ + (f -1)Y + C Z + ε where Zt-j = Yt-j - Yt-j-1 for j = 0, 1, 2, ..., p-1 and εt are white noise processes. So that the process of accepting and rejecting the null hypothesis in the Augmented Dickey-Fuller analysis becomes: 1 1 ˆ 1 ˆ ˆ( ) ADF se      1 ˆ( )se  is the standard error in  - 1. Rejecting the null hypothesis in the unit root test H0 :  = 1 is rejected if ˆ ADF  smaller than the value of α at a significant level. JURNAL AKUNTANSI, MANAJEMEN DAN EKONOMI, VOL 21, N0 1, 2019, 32-45 Johansen Co-integration Test According to Ikudayisi and Salman (2018) and Janzen Sinay (2014: 10) states that for the cointegration test Johansen's cointegration test is used as follows: Yt = Atyt-1 + ... + Apyt-p + Bxt + εt with yt is se is a vector with 𝑘 non stationary variables I (1), 𝑥𝑡 is a vector with 𝑑 deterministic variables, 𝜀𝑡 is an error vector. The equation 𝑉𝐴𝑅 (𝑝) can also be written as p-1 t t-1 i t-i t t i=1 Δy = y + Γ Δy + Bx + ε Where p p i i j i = 1 j=i+1 = A -I, Γ = - A  For hypothesis testing, trace trace statistics can be used: LRtr(r|k) = k i i = r+1 -T log(1 - λ ) and maximum Eigen value test statistics LRmax(r|r + 1) = -T log(1 – λr+1) = LRtr(r|k) – LRtr(r + 1|k) For r = 0, 1, ..., k - 1 with the hypothesis used is H0 : there are r cointegration equations. At the significance level (1 - 𝛼) 100%, 𝐻0 is accepted if the trace test statistics and maximum Eigen value are smaller than the critical value at 𝛼, or 𝑝 𝑣𝑎𝑙𝑢𝑒 greater than the significance value 𝛼. Model Suitability Test According to Sinay (2014), the model compatibility test to see serial correlation on the residuals uses the Portmanteau test statistic as follows: h ' 1 1 h 0 0 j = 1 ˆ ˆ ˆ ˆQ = T tr(C ) j j C C C    Or * 2 ' 1 1 0 0 1 1 ˆ ˆ ˆ ˆ(C ) h h j j j Q T tr C C C T J       with T ' i t tt = i + 1 1ˆ ˆ ˆC = u u -i T  . This test statistic is distributed 2 * 2 ( ( )) X k h n , with n * state the number of coefficients other than constants in the estimated VAR (p) model. The hypothesis proposed in the model compatibility test is: H0: no serial correlation At the significance level (1 - 𝛼) 100%, 𝐻0 is accepted if p 𝑣𝑎𝑙𝑢𝑒 statistics 𝑄 for each lag besar is greater than the significance value 𝛼. Thus, there is no serial correlation. Information Criteria According to Sinay (2014), the selection of order lags can use the following methods: Akaike Information Criterion (AIC) AIC(p) = log det   2 2 ( ) p u k p T  Schwarz Information Criterion (𝑆𝐶) SC(p) = log det   2 u log(T)pk (p) + T  With -1 u (p) = T T t tt = 1 ˆ ˆu u is the size of the sample and k is the number of endogenous variables. The lag value p is chosen as the p* value which minimizes the information criteria in intervals 1, ..., pmax is observed. The optimum lag is based on the most 𝐴𝐼𝐶 and 𝑆𝐶 values small. Causality Analysis Sinay (2014), in the modeling of the Error Error Correction Model (VECM) analysis of causality aims to see long-term causality and short-run causality. Analysis of the long-term causality relationship between the independent variables to the dependent variable in VECM modeling can be seen in the coefficients of the error correction term (ECT), which is based on the sign and results of the coefficient significance test using the 𝑡 test statistic on ^ ^ ^ ^ JURNAL AKUNTANSI, MANAJEMEN DAN EKONOMI, VOL 21, N0 1, 2019, 32-45 the Ordinary Least Square (OLS) method . Meanwhile, for analysis of short- term causality for each variable can use the Granger causality test. The Granger causality test is based on the Wald test statistic which has chi square distribution or test 𝐹 as an alternative. The hypothesis used is 𝐻0 ∶ There is no Granger causality relationship Forecasting and Structural Analysis According to Sinay (2014: 11), the forecasting and structural analysis of VECM bears similarities to forecasting analysis and structural analysis of the VAR model. In VAR modeling the analysis can use impulse response analysis and variance decomposition. Impulse Response analysis aims to see the effect (influence) of each variable (endogenous) if given shock or impulse (shock). Meanwhile, variance decomposition analysis aims to predict contribution of each variable (percentage variance of each variable) caused by changes in certain variables in a system. Like forecasting analysis in general, to determine the accuracy of the forecast results of a model can use the Mean Absolute Percentage Error (𝑀𝐴𝑃𝐸): n t t t = 1 t Ŷ -Y Y MAPE = x 100% n  RESEARCH METHODS The research methods in this study formulated several hypotheses as follows: Is there an Influence of Capital Adequacy, Efficiency and Liquidity on Rentability in syariah bank that owned by Indonesia government from 2009 to 2017. Based on the hypothesis formulation, three variables can be formed as follows: 1. Capital adequacy: Capital Adequacy Ratio (CAR) 2. Efficiency: Operational Costs for Operational Income (OCOI) 3. Liquidity: Financing to Deposit Ratio (FDR) 4. Rentability: Return on Assets (ROA) So that from this study is a case study to analyze the effect of CAR, OCOI and FDR on ROA in the syariah bank that owned by Indonesia bank from 2009 - 2017. Based on the variables that have been formed, then for data sourced from Bank Mandiri Syariah financial statements, Bank BNI Syariah , BRI Bank Syariah and Bank BTN Syariah from 2009-2017. The method used in this study is the VECM (Vector Error Correction Model) method which aims to determine the shape of the four variables above. According to Lexy Janzen Sinay (2014: 12), the procedures in the VECM analysis are as follows: 1. Specifications estimation, and model inspection (Unit root test (sterilization check); Johansen's cointegration test; Model 2. Estimation and Examination 3. Causality analysis 4. Forecasting and structural analysis The results of data processing carried out in this study using EViews 9 software. RESULTS AND DISCUSSION Specifications Estimation And Model Inspection The first step in conducting the unit root test of the four variables includes data from the government of Indonesia ROA, CAR, OEOI and FDR from 2009-2017. Based on data processing using EViews 9, the output results are as follows: Table 1. Unit Root Test (Augmented Dickey- Fuller (ADF) test statistic) Data CV Level 1st Difference (α) Stat. ADF p value Stat. ADF p value ROA - 2,312365 0,1744 - 10,99046 0,0000 5% - 2,957110 - 2,957110 CAR - 0,948478 0,7587 - 11,44974 0,0000 5% - 2,960411 - 2,957110 OEOI - 2,739522 0,0786 - 7,880309 0,0000 5% - 2,957110 - 2,957110 FDR - 1,153066 0,6821 - 13,22417 0,0000 5% - 2,957110 - 2,957110 Based on Table1, an analysis can be made that the data on ROA, CAR, OEOI and FDR in the Indonesian government's Syariah Banks from 2009-2017 are data that contain unit roots at the level or not stationary at the level. This can be seen from the unit root test technique that is done, namely the level technique. It is seen that the ADF statistic JURNAL AKUNTANSI, MANAJEMEN DAN EKONOMI, VOL 21, N0 1, 2019, 32-45 value for each variable is greater than α = 5%. This means accepting the hypothesis H0, namely there is a unit root in data or data that is not stationary. Meanwhile, from the results of the first differentiation, it can be seen in the ADF statistic value of each variable smaller than α = 5%, this means rejecting the hypothesis yaitu0, ie data does not contain unit roots or is stationary. Thus, the variables ROA, CAR, OEOI and FDR are non variables first-order stationer. Johansen Cointegration Test The cointegration test results using lag 2 (significant lag based on the VAR procedure) from the variables ROA, CAR, OEOI and FDR using static trace and maximum Eigen value statistics can be seen in table 6 and table 7. In table 6 it can be seen that the test results hypothesis by using trace statistics for the hypothesis: H0: There is no cointegration connection. P value is 0.0000 smaller than α = 5% (Trace statistic value that is 89.93052 is greater than the value of 47.85613 tables at α = 5%). This means that the hypothesis H0 is rejected. Thus, it can be concluded that there is a cointegration equation. For this reason, the next hypothesis is examined. Table 2. Johansen Cointegration Test H Eigen Value Trace Statistic CV α = 5% p value 0 0,776183 89,93052 47,85613 0,0000 1 0,546901 40,53195 42,79707 0,0620 2 0,311899 14,40770 15,49471 0,0724 3 0,060848 2,071675 3,841466 0,1501 Based on Table 2, the following hypothesis test results will be examined: H0: There is a cointegration equation H1: There is no cointegration equation In table 2, it can be seen that the p-value for each hypothesis is 0,000 smaller than the value of α = 5% (trace statistics greater than the critical value at α = 5% for each hypothesis). This means that H0 is accepted. Thus, based on the analysis it can be concluded that the results of the cointegration test using trace statistics indicate that at least one cointegration equation can be formed. Table 3. Johansen Cointegration Test (Maximum) H Eigen Value Trace Statistic CV α = 5% p value 0 0,776183 49,39857 27,58434 0,0000 1 0,546901 26,12425 28,13162 0,0791 2 0,311899 12,33603 14,26460 0,0987 3 0,060848 2,071675 3,841466 0,1501 From the data contained in table 3 it can be seen that the results of hypothesis testing using maximum Eigen value statistics, namely p-value trace statistics for each hypothesis: H0: There is a cointegration equation and H1: There is no cointegration equation. The p-value in table 7 shows that there is one cointegration, that is, the p-value of 0.0000 is greater than the value of α = 0.05; this means that H0 is accepted. a. Model Estimation and Examination After conducting a cointegration analysis, it is continued by analyzing the optimum lag selection. The optimum lag selection in VECM can use information criteria, namely Akaike Information Criterion (AIC) and Schwarz Information Criterion (𝑆𝐶). The results of data processing using Akaike Information Criterion (AIC) and Schwarz Information Criterion (𝑆𝐶) analysis for one to eight lags can be seen in table 8. Please note that the use of lags one to eight is due to the principle of parsimony (simplicity of models) in statistical modeling, this is caused by the more lag used, the more the parameter parameters of the model. Table 4. Information Criteria Lag Akaike Information Criterion (AIC) Schwarz Information Criterion (𝑆𝐶) 1 20,02776 20,91653 2 18,04871 19,29337* 3 16,29051* 19,66485 4 16,93523 19,40520 In table 4 it can be seen that lag 3 has the smallest AIC value, while lag 2 has the smallest SC value. Thus lag 2 and lag 3 will be used to process the parameter estimation Vector Error Correction Model (VECM). Based on the results of the optimum lag analysis, the VECM equation forms that are estimated are VECM (2) and VECM (3), each with the number of cointegration equations is two. Then the model is examined by selecting the best model between VECM (2) and JURNAL AKUNTANSI, MANAJEMEN DAN EKONOMI, VOL 21, N0 1, 2019, 32-45 VECM (3). Examination of the model is done by using a residual assumption test analysis of the two models, namely the residual serial correlation test as shown in the following table 5: Table 5. Portmanteau Test on VECM (2) and VECM (3) Lg VECM(2) VECM(3) Stat. Q p val db Stat. Q p val db 1 16,46283 NA* NA* 15,39711 NA* NA* 2 36,09632 NA* NA* 33,28613 NA* NA* 3 54,29685 0,1345 16 47,66786 NA* NA* 4 73,48157 0,1564 32 62,66859 0,0165 16 5 86,86297 0,2165 48 78,70373 0,0187 32 6 104,6527 0,2453 64 100,3596 0,0265 48 7 110,1408 0,2675 80 115,5666 0,0365 64 8 124,3054 0,3123 96 136,2712 0,0386 80 9 135,0840 0,3453 112 145,5644 0,0653 96 10 146,2771 0,3564 128 154,4738 0,0754 112 11 156,3912 0,3675 144 165,0873 0,0875 128 12 164,7442 0,3876 160 177,4790 0,0894 144 In table 5, it can be seen that the Portmanteau test results for VECM (2) do not contain serial residual correlation at each lag. Whereas for VECM (3) states that the model contains serial correlation residual in lag 4,5,6,7,8, where lag 3 p-value of Q-statistic for lag is less than significance level α = 5% (meaning reject H0 : no serial correlation). Thus, VECM (2) is better than VECM (3) because there are no residual serial correlations. This means that VECM (2) is the best model. ΔROAt = –1,692(ROAt-1 – 0,075CARt-1 + 0,092OEOI t-1 – 0,040FDR t-1 – 4,496) – 0,308ΔROAt-1 – 0,127ΔROAt-2 + 0,071ΔCARt-1 + 0,095ΔCARt-2 + 0,007ΔOEOIt-1 + 0,020ΔOEOIt-2 + 0,045ΔFDRt-1 + 0,064ΔFDRt-2 + 0,082 (1) ΔCARt = –1,900(ROA t-1 – 0,075CARt-1 + 0,092OEOIt-1 – 0,040FDRt-1 – 4,496) – 8,562ΔROAt-1 – 4,909ΔROAt-2 – 1,056ΔCARt-1 – 0,318ΔCARt-2 – 0,909ΔOEOIt-1 – 0,436ΔOEOIt-2 – 0,145ΔFDRt-1 – 0,149ΔFDRt-2 – 0,198 (2) ΔOEOIt = –27,912(ROAt-1 – 0,075CARt-1 + 0,092OEOIt-1 – 0,040FDRt-1 – 4,496) + 1,357ΔROAt-1 + 2,230ΔROAt-2 – 0,905ΔCARt-1 – 0,805ΔCARt-2 – 0,237ΔOEOIt-1 – 0,096ΔOEOIt-2 – 0,776ΔFDRt-1 – 0,743ΔFDRt-2 – 1,077 (3) ΔFDRt = –33,288(ROAt-1 – 0,075CARt-1 + 0,092OEOIt-1 – 0,040FDRt-1 4,496) – 8,166ΔROAt-1 + 7,830ΔROAt-2 + 3,320ΔCARt-1 + 2,288ΔCARt-2 – 0,059ΔOEOIt-1 + 0,834ΔOEOIt-2 +0,726ΔFDRt-1 + 0,681ΔFDRt-2 + 0,836 (4) Granger Causality Analysis for Dependent Variables AROAt Based on the form of equation (1) of the VECM model (2) above, it is known that the ROA variable has a cointegration equation – 1.692, where the error correction term (ECT) coefficient is negative. Based on the results of data processing using Eviews 9, it was found that for p-value F statistics of 0.000031 less than the significance level α = 5%, which means that the coefficient is significant. Thus, the coefficient of ECT in equation (1) is a significant coefficient and is negative. This means that there is a long-run causality from CAR, OEOI and FDR to the level of ROA in Islamic banks owned by the Indonesian government from 2009-2017. This result is supported by research conducted by Mawardi (2005) which states that CAR has a long-term causality relationship to ROA. According to Lexy Janzen Sinay (2014: 14) to see the relationship of short-run causality in a VECM equation, the Granger causality test was used. The results of the Granger causality test in the first equation of VECM (2) are as follows: Table 6. Granger Causality Test: Dependent Variable ΔROAt Excluded Chi-sq db p value ΔCARt 9,512895 2 0,0086 ΔOEOIt 0,253455 2 0,8810 ΔFDRt 28,72819 2 0,0000 All 57,61088 6 0,0000 In table 9, an analysis can be made, that is Wald's p-value statistical test. Variable ΔCARt is 0.0086 smaller than the significant level α = 5%. This means that, rejecting hypothesis H0: there is no short- term causality relationship which means that there is a short-term causality relationship between the level of CAR and the level of JURNAL AKUNTANSI, MANAJEMEN DAN EKONOMI, VOL 21, N0 1, 2019, 32-45 ROA in syariah bank that owned by government of Indonesia from 2009-2017. For the variable ΔOEOIt has a p-value of 0.8810 greater than the significance level α = 5%. This means that, accepting hypothesis H0: there is no short-term causality relationship which means there is no short- term causality relationship between the OEOI level and the level of ROA in syariah bank that owned by government of Indonesia from 2009-2017. Next for the ΔFDRt variable has a p-value of 0.0000 less than the significant level α = 5%. This means that, rejecting hypothesis H0: there is no short-term causality relationship which means there is a short-term causality relationship between the FDR level to the level of ROA in syariah bank that owned by government of Indonesia from 2009-2017. However, if viewed as a whole in equation (1), then there is a short-term causality relationship from the level of CAR, OEOI and FDR to the level of ROA in Indonesian government-owned Islamic Banks from 2009- 2017. This can be seen from the value of p- value = 0.0000 smaller than the significant level α = 5%, which means rejecting the hypothesis H0 where this means there is a short-term causality relationship between the levels of CAR, OEOI and FDR to the level of ROA in Islamic Banks owned by the Indonesian government from 2009-2017. This is consistent with the research conducted by Anita Karisma Mastika Permatasari and Dheasey Amboningtyas (2017). Granger Causality Analysis for Dependent Variables ΔCARt Based on the form of equation (2) of the VECM (2) model above, it is known that the CAR variable has a cointegration equation – 1,900; where the error correction term (ECT) coefficient is negative. Based on the results of data processing using Eviews 9, it was found that for p-value F statistics of 0,000001 less than the significance level α = 5%, which means that the coefficient is significant. Thus, the coefficient of ECT in equation (2) is a significant coefficient and is negative. This means that there is a long-run causality of ROA, OEOI and FDR against the level of CAR in the Indonesian government-owned Islamic Banks from 2009-2017. This is in accordance with the research conducted by Fenandi Bilian and Purwanto (2017). According to Lexy Janzen Sinay (2014) to see the relationship of short-run causality in a VECM equation, the Granger causality test was used. The results of the Granger causality test in the first equation of VECM (2) are as follows: Table 7. Granger Causality Test: Dependent Variable ΔCARt Excluded Chi-sq db p value ΔROAt 18,23495 2 0,0001 ΔOEOIt 18,09765 2 0,0001 ΔFDRt 6,079053 2 0,0479 All 39,01579 6 0,0000 In table 7, an analysis can be made, namely the Wald p-value statistical test. Variable ΔROAt is 0,0001 smaller than significant level α = 5%. This means that, rejecting hypothesis H0: there is no short- term causality relationship which means that there is a short-term causality relationship between the level of ROA to the level of CAR in syariah bank that owned by government of Indonesia from 2009-2017. For the variable ΔOEOIt has a p-value of 0,0001 smaller than the significant level α = 5%. This means that, rejecting hypothesis H0: there is no short-term causality relationship which means there is a short- term causality relationship between the OEOI level and the level of CAR in syariah bank that owned by government of Indonesia from 2009-2017. Next for the ΔFDRt variable has a p-value of 0.0479 smaller than the significant level α = 5%. This means that, rejecting hypothesis H0: there is no short-term causality relationship which means there is a short- term causality relationship between the FDR level to the level of CAR in syariah bank that owned by government of Indonesia from 2009-2017. However, if viewed as a whole in equation (2), then there is a short-term causality relationship from the level of ROA, OEOI and FDR to the level of CAR in the Indonesian government-owned Islamic Bank from 2009- 2017. This can be seen from the value of p- value = 0.0000 smaller than the significant level α = 5%, which means rejecting the hypothesis H0, which means that there is a short-term causality relationship between the level of ROA, OEOI and FDR to the level of CAR in a Sharia Bank owned by the Indonesian government from 2009-2017. This is supported by research conducted by Rofikoh Rokhim and Jubilant Arda Harmidy (2013). JURNAL AKUNTANSI, MANAJEMEN DAN EKONOMI, VOL 21, N0 1, 2019, 32-45 Granger Causality Analysis for Dependent Variables ΔOEOIt Based on the form of equation (3) of the VECM (2) model above, it is known that the OEOI variable has a cointegration equation – 27,912; where the error correction term (ECT) coefficient is negative. Based on the results of data processing using Eviews 9, it was found that for p-value F statistics of 0,000002 less than the significance level α = 5%, which means that the coefficient is significant. Thus, the ECT coefficient in equation (3) is a significant coefficient and is negative. This means that there is a long-run causality from ROA, CAR and FDR to the level of OEOI in the Indonesian government Islamic banks from 2009-2017. This is in accordance with the research conducted by Deden Edwar Yokeu Bernardin (2016). According to Lexy Janzen Sinay (2014) to see the relationship of short-run causality in a VECM equation, the Granger causality test was used. The results of the Granger causality test in the first equation of VECM (2) are as follows: Table 8. Granger Causality Test: Dependent Variable ΔOEOIt Excluded Chi-sq db p value ΔROAt 0,322428 2 0,8511 ΔCARt 8,221816 2 0,0164 ΔFDRt 40,28180 2 0,0000 All 71,83938 6 0,0000 In table 8, an analysis can be made, namely Wald's p-value statistical test. Variable ΔROAt is 0.8511 greater than the significance level α = 5%. This means that, accepting hypothesis H0: there is no short- term causality relationship which means that there is no short-term causality relationship between the level of ROA to the level of OEOI in syariah bank that owned by government of Indonesia from 2009-2017. For the variable ΔCARt has a p-value of 0.0164 smaller than the significant level α = 5%. This means that, rejecting hypothesis H0: there is no short-term causality relationship which means that there is a short-term causal relationship between the level of CAR and the level of OEOI in syariah bank that owned by government of Indonesia from 2009-2017. Next for the ΔFDRt variable has a p-value of 0,000 smaller than the significant level α = 5%. This means that, rejecting hypothesis H0: there is no short-term causality relationship which means there is a short-term causality relationship between the FDR level to the OEOI level in syariah bank that owned by government of Indonesia from 2009-2017. However, if viewed as a whole in equation (3), then there is a short-term causality relationship from the level of ROA, CAR and FDR to the OEOI level in syariah bank that owned by government of Indonesia from 2009-2017. This can be seen from the value of p-value = 0,0000 smaller than the significant level α = 5%, which means rejecting the hypothesis H0, which means that there is a short-term causality relationship between the level of ROA, CAR and FDR to the OEOI level in syariah bank that owned by government of Indonesia from 2009-2017. Granger Causality Analysis for Dependent Variables DRFDRt Based on the form of equation (4) of the VECM model (2) above, it is known that the FDR variable has a cointegration equation – 33,288; where the error correction term (ECT) coefficient is negative. Based on the results of data processing using Eviews 9, it was found that for p-value F statistics of 0,000000 less than the significance level α = 5%, which means that the coefficient is significant. Thus, the coefficient of ECT in equation (4) is a significant coefficient and is negative. This means that there is a long-run causality from ROA, CAR and OEOI to the level of FDR in Islamic banks owned by the Indonesian government from 2009-2017. This is in accordance with the research conducted by Erma Kurniasih (2016). According to Lexy Janzen Sinay (2014) to see the relationship of short-run causality in a VECM equation, the Granger causality test was used. The results of the Granger causality test in the first equation of VECM (2) are as follows: Table 9. Granger Causality Test: Dependent Variable ΔFDRt Excluded Chi-sq db p value ΔROAt 11,83162 2 0,0027 ΔCARt 37,87209 2 0,0000 ΔOEOIt 4,282536 2 0,1175 All 101,1076 6 0,0000 In table 9, an analysis can be made, that is Wald's p-value statistical test. Variable ΔROAt is 0.0027 smaller than the significant level α = 5%. This means that, rejecting hypothesis H0: there is no short- term causality relationship which means that there is a short-term causality relationship between the level of ROA to the FDR level in JURNAL AKUNTANSI, MANAJEMEN DAN EKONOMI, VOL 21, N0 1, 2019, 32-45 syariah bank that owned by government of Indonesia from 2009-2017. For the variable ΔCARt has a p-value of 0.0000 smaller than the significant level α = 5%. This means that, rejecting hypothesis H0: there is no short-term causality relationship which means there is a short- term causality relationship between the level of CAR to the FDR level in syariah bank that owned by government of Indonesia from 2009-2017. Next for the variable OPOEOIt has a p-value of 0.1175 greater than the significant level α = 5%. This means that, accepting the hypothesis H0: there is no short-term causality relationship which means there is no short-term causality relationship between the OEOI level of the FDR level in syariah bank that owned by government of Indonesia from 2009-2017. However, if we look at it as a whole in equation (4), there is a short-term causality relationship from the level of ROA, CAR and OEOI to the FDR level in the government- owned Islamic Bank from 2009-2017. This can be seen from the value of p-value = 0,0000 smaller than the significant level α = 5%, which means rejecting the hypothesis H0, which means that there is a short-term causality relationship between the level of ROA, CAR and OEOI on the FDR level in Islamic Banks owned by the Indonesian government from 2009-2017. This is consistent with the research conducted by Kamalia Sani and Maftukhatusolikhah (2015). Forecasting and Structural Analysis This section will explain about forecasting and structural analysis of forecasting from the VECM model (2). Before discussing the analysis of forecasting results it will be explained in advance about structural analysis which includes the Impulse Response Function (IRF) analysis and variance decomposition. Impulse Response Function (IRF) analysis is the result of impulse-response (IRF) plots where there are nine Impulse Response Function (IRF) plots for the next 10 periods, which explain visually the response (response) of a variable that arises due to shocks ( shock / impulse) of one standard deviation both from itself and from other variables. Impulse Response Function (IRF) Analysis The Impulse Response Function (IRF) analysis can be seen from the following picture: From the data shown in Figure 1, it can be seen that the average response to the shocks of each variable is stagnant both positive and negative. For those who have a positive stagnant pattern such as the ROA response to itself which is in a positive area. And for the negative stagnant pattern as possessed by the OEOI level response to ROA that has a negative stagnant pattern. Figure 1. Impulse Response Function (IRF) Analysis JURNAL AKUNTANSI, MANAJEMEN DAN EKONOMI, VOL 21, N0 1, 2019, 32-45 Forecast Error Decomposition Variance (FEDV) Analysis of variance decomposition is often referred to as a forecast error decomposition variance (FEDV) analysis. The results of the FEDV analysis for the 10 periods of each variable are as follows table 10: Table 10. Forecast Error Decomposition Variance (FEDV) for Variable ROA Period S.E. ROA CAR OEOI FDR 1 0,558821 100,0000 0,000000 0,000000 0,000000 2 0,811845 84,65184 3,384199 9,065921 2,898036 3 0,955230 83,07784 4,692889 7,982179 4,247092 4 1,013536 81,77571 4,203757 8,308055 5,712473 5 1,117603 84,10387 3,497573 7,541955 4,856599 6 1,201076 83,38625 4,068939 7,508163 5,036644 7 1,308588 84,01748 3,816091 7,547739 4,618685 8 1,356675 84,19319 3,571419 7,586180 4,649215 9 1,434692 84,87481 3,288029 7,623153 4,214004 10 1,490787 84,75780 3,365720 7,440078 4,436406 Table 10 is a summary of the results of the FEDV analysis for the ROA level of shocks given by each variable including itself. The FEDV analysis that can be taken from table 14 states that in the short term, the third period: shocks to itself result in 83,08% fluctuations in the level of ROA, and shocks to the level of CAR result in 4,69% fluctuations in the ROA level, while the OEOI level results in 7,99% fluctuations in the level of ROA and the FDR rate resulted in 4,25% of the fluctuations in the level of ROA in syariah bank that owned by Indonesia government from 2009-2017. On the other hand, in the long term, on the 10 th period; the fluctuations in themselves increased in the ROA level, while the CAR level shock weakened by 3,37% in the ROA level, while the OEOI level weakened by 7,4% in the ROA level and the FDR rate increased by 4.4% in ROA level. Then for FEDV analysis for the level of CAR can be seen in the following table 11 : Table 11. Forecast Error Decomposition Variance (FEDV) for Variable CAR Peroid S.E. ROA CAR OEOI FDR 1 2,569470 0,216592 99,78341 0,000000 0,000000 2 3,265500 11,06809 61,77981 10,30841 16,84369 3 3,430043 10,58650 57,42224 10,13297 21,85829 4 3,543078 10,62545 55,18360 12,20149 21,98945 5 3,739761 9,555302 59,37204 10,98054 20,09211 6 3,965661 11,24582 55,34150 12,01947 21,39321 7 4,108551 10,53188 53,92738 11,19986 24,34088 8 4,229213 10,53545 52,02640 11,91639 25,52176 9 4,321707 10,11199 53,21140 11,44703 25,22958 10 4,457600 10,44308 52,69517 11,51112 25,35062 Table 11 is a summary of the results of the FEDV analysis for the level of CAR from shocks given by each variable including itself. The FEDV analysis that can be taken from table 15 states that in the short term the third period: the shock to itself results in 10,59% fluctuations in the level of CAR, and shocks to itself result in 57,42% fluctuations in the level of CAR, while the OEOI level 10,13% fluctuations in the level of CAR and the FDR level resulted in 21,86% of the fluctuations in the level of CAR in syariah bank that owned by Indonesia government from 2009-2017. On the other hand, in the long term, on the 10 th period; the shock to itself increased its fluctuation in the ROA level of 10,44 in the CAR level, while the CAR level shock against itself weakened by 52,70% in the CAR level, while the OEOI level increased by 11,51% in the CAR level and the FDR rate rose by 25,35% in the CAR level. Then for analysis of FEDV for OEOI level can be seen in the following table 12 : JURNAL AKUNTANSI, MANAJEMEN DAN EKONOMI, VOL 21, N0 1, 2019, 32-45 Table 12. Forecast Error Decomposition Variance (FEDV) for Variable OEOI Period S.E. ROA CAR OEOI FDR 1 4,900021 90,65639 0,256969 9,086645 0,000000 2 8,543111 68,82259 12,25162 13,09900 5,826784 3 9,734143 73,24532 10,39652 11,46934 4,888820 4 10,33510 72,75164 9,542467 10,57479 7,131103 5 11,27413 75,99279 8,237227 9,144431 6,625551 6 12,21335 75,72465 8,966293 8,544613 6,764440 7 13,11041 78,03873 7,804636 8,281893 5,874737 8 13,62512 78,34765 7,481587 7,928540 6,242224 9 14,35411 79,53221 6,808952 7,645938 6,012897 10 14,97808 79,48568 6,961552 7,269908 6,282861 Table 12 is a summary of the results of the FEDV analysis for the OEOI level of shocks given by each variable including itself. The FEDV analysis that can be taken from table 16 states that in the short term the third period: shocks to itself result in 11,47% fluctuations in the OEOI level, and shocks to ROA result in 73,26% fluctuations in the ROA level, while the CAR level results in 10,40% of fluctuations in the OEOI level and FDR rate resulted in 4.89% of the fluctuations in the OEOI level in bank syariah that owned by Indonesia government from 2009-2017. On the other hand, in the long term, on the 10th period; self-shock has increased fluctuations in the ROA level of 79,49 in the OEOI level, while the OEOI level shock against itself has weakened by 7,27% in the OEOI level, while the CAR level has weakened to 6,96% in the OEOI level and the FDR rate weakened to 6,28% at the OEOI level. Then for FEDV analysis for FDR levels can be seen in the following table 13 : Table 13. Forecast Error Decomposition Variance (FEDV) for Variable FDR Period S.E. ROA CAR OEOI FDR 1 7,217440 25,40206 4,078983 0,049633 70,46932 2 8,914190 19,03179 3,518098 24,05266 53,39746 3 9,624438 17,04577 10,92353 20,66370 51,36700 4 10,04115 15,68744 15,23107 19,99362 49,08787 5 11,39334 21,15253 11,88123 16,33653 50,62972 6 11,94191 20,11164 10,84064 17,99694 51,05078 7 12,79217 19,17657 12,93247 16,78297 51,10799 8 12,88804 18,89356 13,74062 16,97900 50,38681 9 13,51641 21,38142 12,49306 16,61052 49,51499 10 13,82055 21,10662 12,24002 16,70587 49,94748 Table 13 is a summary of the results of the FEDV analysis for the FDR level of shocks given by each variable including itself. The FEDV analysis that can be taken from table 17 states that in the short term the third period: shocks to itself result in 51,37% fluctuations in the FDR level, and shocks to ROA result in 17,05% fluctuations in the FDR level, while the CAR level results in 10,92% of fluctuations in FDR and OEOI levels resulted in 20,66% of fluctuations in FDR levels in the Indonesian government-owned Islamic banks from 2009-2017. On the other hand, in the long term, on the 10th period; the shock to itself increased its fluctuation in the ROA level to 21,11% in the FDR level, while the FDR level shock to itself weakened to 49,95% in the FDR level, while the CAR level increased to 12,24% at the level FDR and OEOI levels weakened to 16,71% in the FDR level. JURNAL AKUNTANSI, MANAJEMEN DAN EKONOMI, VOL 21, N0 1, 2019, 32-45 Forecast Result The forecast results using VECM (2) for the next 10 periods can be seen in the following figure 2 : Figure 2. Prediction results for ROA, CAR, OEOI and FDR From the figure above, it is known that it is predicted that there will be an increase in the level of ROA while the one that experienced a decline is the level of OEOI, besides that it will experience stagnant growth, namely the level of CAR and FDR in the bank syariah owned by Indonesia government from 2009- 2017. Based on Figure 2, MAPE values can be obtained from each variable as shown in the following table 14 : Table 14. Accuracy of forecast results ROA CAR OEOI FDR MA PE 99,98 % 100,9 1% 100,3 8% 3185,3 4% Based on Figure 17, MAPE values can be obtained from each variable as shown in table 18. In table 18 it can be seen that the smallest MAPE is the ROA variable. This means that forecasting using the VECM model (2) is more accurate if applied to ROA (Return On Asset). CONCLUSION Based on the results and discussion, conclusions can be made as follows. Based on the model specifications (optimum lag analysis) and model checking (residual serial correlation test), the best model for data on CAR, OEOI and FDR to ROA is obtained is VECM (2). Based on VECM (2), the results of causality analysis are obtained as follows. There is a short-term and long-term causality relationship between the level of ROA as the dependent variable and the level of CAR, OEOI and FDR as independent variables. There is no short-term causality relationship between the level of ROA as the dependent variable with the level of CAR, OEOI and FDR, but conversely there is a long-term relationship between the level of ROA to the level of CAR, OEOI and FDR. There is no relationship between short and long term causality between the level of ROA as the dependent variable with the level of CAR, OEOI and FDR. Based on the structural analysis of VECM (2), it can be concluded that the response of each variable to the shock that comes from itself is quite significant, due to fluctuations. The response of the level of ROA to CAR, OEOI and FDR is very significant. In general, for future analysis both in the long and the short term, the level of ROA on the levels of CAR, OEOI and FDR significantly influence each other. The forecast results obtained using VECM (2) are quite accurate, especially for forecasting ROA levels. This can be seen from the MAPE of the level of ROA which shows the smallest percentage number of the four variables. It is important to know that the level of ROA in Islamic banks owned by the government of Indonesia shows the level of profits achieved by banks in a certain period that are sourced from the total assets they have. REFERENCES Bernardin, D.E.Y. (2016). Pengaruh CAR dan LDR terhadap beban operasional pendapatan operasional. Jurnal Ecodemica, 4(2):232- 241. Bilian, B. & Purwanto. (2017). Analisis pengaruh CAR, NIM, BOPO dan LDR terhadap profitabilitas bank persero. Firm Journal of Management Studies, 2(1):155-168. JURNAL AKUNTANSI, MANAJEMEN DAN EKONOMI, VOL 21, N0 1, 2019, 32-45 Fadjar, A., Hedwigis E.R. & Prihatini, T. (2013). Analisis faktor internal dan eksternal bank yang mempengaruhi profitabilitas bank umum di Indonesia. Journal of Management and Business Review, 10(1):63-67. Faradila, F. & Cahyati A.D. (2013). Analisis manajemen laba pada perbankan syariah. Jurnal Riset Akuntansi & Komputerisasi Akuntansi, 4(1):57-74. Hamilton, J. D. (1990). Analysis of time series subject to changes in regime. Journal of Econometrics, (45) : 39-70. Giannini, N.G. (2013). Faktor yang mempengaruhi pembiayaan Mudharabah pada bank umum syariah di Indonesia. Accounting Analysis Journa,l 2(1):97-103. Ikudayisi, A.A & Salman, K.K. (2018). Spatial integration of maize market in Nigeria – A vector error correction model. International Journal of Food and Agricultural Economics, 2(3):71-80. Kurniasih, K. 2016. Pengaruh capital adequacy ratio, non performing loan, loan to deposit ratio, efisiensi operasi, not interest margin terhadap finance deposit ratio (Studi empiris pada perusahaan perbankan yang listing di BEI tahun 2009–2014). Journal of Accounting, 2(2):23-35. Moroke, N.D. Petersen, J,M. & Petersen, M. (2014). A multivariate time series analysis of household debts during 2007-2009 financial crisis in South Africa: A vector error correction approach. Mediterranean Journal of Social Sciences, 5(7):107-118. Nazrian, A. & Hidayat, P. (2012). Studi tentang keputusan nasabah dalam menabung di Bank Sumut cabang USU Medan metode analytical hierarchy process (AHP). Jurnal Ekonomi dan Keuangan, 1(1):14-21. Permatasari, A.K.M & Amboningtyas, D. (2017). The influence of CAR, BOPO and FDR on ROA (study on conventional bank sub sector company 2012-2016 listed in BEI). Journal of Management, 3(3):21-34. Rokhim, R. & Harmidy, J.A. 2013. Ownership structure, risk and their impact towards performances in Indonesian syariah bank. Finance and Banking Journal. 15(1):22-37 Sani, K. & Maftukhatusolikhah. (2015). Pengaruh capital adequacy ratio (CAR), BOPO, return on asset terhadap finance deposit ratio pada bank umum syariah di Indonesia 2011-2013. I-Economics Journal, 1(1):34-45. Sari, N. (2016). Manajemen dana bank syariah. Jurnal Ilmu Syariah:Al-Maslahah, 12(1):45-61 Sinay, L.J. (2014). Vector error correction model approach to analysis of the relationship of inflation, BI Rate and US Dollar. Jurnal Barekeng, 8(2):9-18. Wahyu, D.R. 2016. Financing to deposit ratio (FDR) sebagai salah satu penilaian kesehatan bank umum syariah (Study kasus pada Bank BJB Syariah cabang Serang). Jurnal Ekonomi Keuangan dan Bisnis Islam, 7(1):19-36. Wibisono, M.F. & Wahyuni, S. (2017). Pengaruh CAR, NPF, BOPO, FDR terhadap ROA yang dimediasi oleh NOM. Journal of Business and Management, 17(1):41-62 Wisnu, M. 2005. Analisis faktor-faktor yang mempengaruhi kinerja keuangan bank umum di indonesia (studi kasus pada bank umum dengan total assets kurang dari 1 trilliun). Jurnal Bisnis Strategi, 3(3):84-94 .